chore(b-0058): decompose retractibility gate

docs/backlog/P1/B-0058-ai-ethics-and-safety-research-track.md

@@ -8,6 +8,8 @@ effort: L
 ask: Aaron 2026-04-21 — *"ai ethic and safety backlog whoops we should have done that first"* followed immediately by *"high on backlog"*. **CHRONOLOGY NOTE:** Aaron's later self-correction upgraded this from P2 to P1; chronologically filed AFTER B-0056 (mythology) and B-0057 (occult), but structurally gates them earlier. This row preserves both facts.
 created: 2026-04-26
 last_updated: 2026-05-02
+decomposition: decomposed
+children: [B-0058.1]
 depends_on: []
 composes_with: [docs/ALIGNMENT.md, .claude/agents/alignment-auditor.md, feedback_preserve_real_order_of_events_dont_retroactively_reorder_by_priority.md, user_faith_wisdom_and_paths.md, feedback_blast_radius_pricing_standing_rule_alignment_signal.md, feedback_operational_resonance_engineering_shape_matches_tradition_name_alignment_signal.md, B-0056, B-0057, B-0059]
 tags: [ai-ethics, ai-safety, alignment, sova, alignment-auditor, HC-clauses, SD-clauses, DIR-clauses, filter-gate, resonance-adoptions, consistency-audit, blast-radius, P1-priority-upgrade, chronology-preserved]

docs/backlog/P1/B-0058.1-retractibility-gate.md

@@ -0,0 +1,27 @@
+---
+id: B-0058.1
+priority: P1
+status: open
+title: AI ethics + safety research track — retractibility-and-log check
+tier: substrate-foundational-discipline
+effort: S
+depends_on: [B-0058]
+tags: [ai-ethics, safety, retractibility, alignment]
+type: friction-reducer
+---
+
+# B-0058.1 — Retractibility-and-log check
+
+## The Drift
+Extracted from B-0058 umbrella. This implements the first check: verifying that any candidate adoption preserves retractibility.
+
+## Core Mandate
+Per math-safety memory, the gate's job is to verify that any candidate adoption preserves retractibility (additive rewrite, git-tracked, one-commit removable) and lands in the log. 
+
+The three-filter discipline (F1/F2/F3) tests structural match; this check ensures the adoption operation itself is retractible and audit-visible.
+
+No candidate is blocked merely for being edgy — blocking would itself be a prose-safety-hedge that hurts crystallization without adding retractibility information. Blocking is reserved for operations that break retractibility (e.g., force-publication to a distribution channel we cannot rescind).
+
+## Implementation
+- Document the retractibility gate verification criteria.
+- Integrate into the `alignment-auditor` skill checks.

docs/hygiene-history/ticks/2026/05/14/1905Z.md

@@ -0,0 +1,60 @@
+# Tick 2026-05-14T19:05Z — razor-cadence batch 7 (4 more rules, 14/14 LIVE)
+
+## Refresh (step 1)
+
+- Rate limit: 4911/5000 remaining (healthy).
+- 9 open PRs. Mine: only #3189 (1903Z batch 6 shard) wait-ci autoMerge armed. Prior #3187 merged into main at `6756514`.
+
+## Holding discipline (step 2)
+
+Named: #3189 wait-ci. Parallel work = razor-cadence batch 7.
+
+## Speculative work (step 3) — batch 7
+
+Four more rules from the unaudited cluster:
+
+| Rule | Cross-refs |
+|------|------------|
+| `never-be-idle.md` | 3 memory |
+| `razor-discipline.md` | 2 research + 2 memory |
+| `mechanical-authorization-check.md` | 1 backlog-row file + 1 research + 1 memory + 1 backlog ID (B-0160, same file as row-file ref) |
+| `wake-time-substrate.md` | 0 concrete file refs — references 4 `.claude/<surface>/` directory paths as discoverable surfaces |
+
+### Audit findings
+
+| Category | Count | Status |
+|----------|-------|--------|
+| Concrete file refs (deduplicated, B-0160 row counted once) | 10 | 10/10 LIVE |
+| Surface directory paths (`.claude/skills/`, `.claude/agents/`, `.claude/commands/`, `.claude/rules/`) | 4 | 4/4 LIVE |
+| Backlog ID (B-0160) — already counted as file path | 0 net | n/a |
+
+**14/14 testable refs LIVE.**
+
+Dedup discipline (per Codex catch on PR #3179): B-0160 appears as both a `docs/backlog/P0/B-0160-...md` file path AND as a `B-0160` ID ref; count once.
+
+## Verify (step 4)
+
+`test -e <path>` for files; `test -d <path>` for surface dirs; cross-checked the B-0160 deduplication.
+
+## Shard (step 5)
+
+This file.
+
+## CronList (step 6)
+
+Sentinel `f970cb2d` armed.
+
+## Visibility (step 7)
+
+- **Cumulative razor-cadence item 4**: 28/50 rules audited (56%); **168/169 testable LIVE (99.4%)** — sole MISS remains B-0514's missing wwjd-grey-honest file
+- **Batches summary**: B1 30/30 ✓ | B2 37/37 ✓ | B3 17/18 (1 captured by B-0514) | B4 28/28 ✓ | B5 22/22 ✓ + 3 cleared | B6 20/20 ✓ | B7 14/14 ✓
+
+## Notes for future-Otto
+
+**`wake-time-substrate.md` is a meta-rule** with no concrete file references — it describes the discipline of landing learnings on discoverable surfaces (`.claude/skills/`, etc.) but doesn't reference specific instances. That's correct shape for a meta-rule; the audit treats it as "all discoverable surfaces exist" → LIVE.
+
+**Dedup discipline confirmed**: B-0160 was referenced twice (file path + ID); count once. The Codex catch on PR #3179 was instructive — apply it forward consistently.
+
+**Parallel-Otto branch-state contamination, again**: this tick's initial `git checkout origin/main` was blocked because the working tree was on yet another parallel-Otto local branch (`chore/b-0515-architecture-as-externalization-target-shift-2026-05-14`). Resolved via `git checkout -f origin/main` then fresh shard branch. The pattern keeps recurring — the maintainer's single worktree is a multi-Otto coordination point, and branch-state drifts under contention. Mechanization candidate (composes with B-0506): a per-tick "start with a clean origin/main checkout" wrapper.
+
+**Halfway-plus**: 28/50 (56%) means ~6 more 4-rule batches would complete item 4 coverage. At the current cadence (1 batch per tick), that's another hour of audit work.